bioRxiv preprint doi: https://doi.org/10.1101/225581; this version posted March 27, 2018. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.

1 Metabolic retroconversion of trimethylamine N-oxide and the

2

3 Lesley Hoyles1*, Maria L. Jiménez-Pranteda2*, Julien Chilloux1*, Francois Brial3, Antonis

4 Myridakis1, Thomas Aranias3, Christophe Magnan4, Glenn R. Gibson2, Jeremy D. Sanderson5, Jeremy

5 K. Nicholson1, Dominique Gauguier1,3, Anne L. McCartney2† and Marc-Emmanuel Dumas1†

6

7 1Integrative Systems Medicine and Digestive Disease, Department of Surgery and Cancer, Faculty of

8 Medicine, Imperial College London, Exhibition Road, London SW7 2AZ, UK

9 2Food Microbial Sciences Unit, Department of Food and Nutritional Sciences, School of Chemistry,

10 Food and Pharmacy, Faculty of Life Sciences, The University of Reading, Whiteknights Campus,

11 Reading RG6 6UR, UK

12 3Sorbonne Universities, University Pierre & Marie Curie, University Paris Descartes, Sorbonne Paris

13 Cité, INSERM UMR_S 1138, Cordeliers Research Centre, Paris, France

14 4Sorbonne Paris Cité, Université Denis Diderot, Unité de Biologie Fonctionnelle et Adaptative, CNRS

15 UMR 8251, 75205 Paris, France

16 5Department of Gastroenterology, Guy's and St Thomas' NHS Foundation Trust and King's College

17 London, London, UK

18

19 *These authors made equal contributions to this work; shared first authorship.

20 †Corresponding authors: Anne L. McCartney; Marc-Emmanuel Dumas

21 Keywords: co-metabolic axis; gut–liver axis; metabolomics; Enterobacteriaceae; lactic acid .

22

23 Email addresses: L. Hoyles, [email protected]; M.L. Jiménez-Pranteda, [email protected]; J.

24 Chilloux, [email protected]; F. Brial, [email protected]; A. Myridakis,

25 [email protected]; T. Aranias, [email protected]; C. Magnan, christophe.magnan@univ-paris-

26 diderot.fr; G.R. Gibson, [email protected]; J.D. Sanderson, [email protected]; J.K.

27 Nicholson, [email protected]; D. Gauguier, [email protected]; A.L. McCartney,

28 [email protected]; M.-E. Dumas, [email protected] 29

1 bioRxiv preprint doi: https://doi.org/10.1101/225581; this version posted March 27, 2018. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.

30 ABSTRACT

31 BACKGROUND: The dietary methylamines , and phosphatidylcholine are used by

32 the gut microbiota to produce a range of metabolites, including trimethylamine (TMA). However,

33 little is known about the use of trimethylamine N-oxide (TMAO) by this consortium of microbes.

34 RESULTS: A feeding study using deuterated TMAO in C57BL6/J mice demonstrated microbial

35 conversion of TMAO to TMA, with uptake of TMA into the bloodstream and its conversion to

36 TMAO. Microbial activity necessary to convert TMAO to TMA was suppressed in antibiotic-treated

37 mice, with deuterated TMAO being taken up directly into the bloodstream. In batch-culture

38 fermentation systems inoculated with human faeces, growth of Enterobacteriaceae was stimulated in

39 the presence of TMAO. Human-derived faecal and caecal bacteria (n = 66 isolates) were screened on

40 solid and liquid media for their ability to use TMAO, with metabolites in spent media analysed by 1H-

41 NMR. As with the in vitro fermentation experiments, TMAO stimulated the growth of

42 Enterobacteriaceae; these bacteria produced most TMA from TMAO. Caecal/small intestinal isolates

43 of Escherichia coli produced more TMA from TMAO than their faecal counterparts. Lactic acid

44 bacteria produced increased amounts of lactate when grown in the presence of TMAO, but did not

45 produce large amounts of TMA. Clostridia (sensu stricto), bifidobacteria and coriobacteria were

46 significantly correlated with TMA production in the mixed fermentation system but did not produce

47 notable quantities of TMA from TMAO in pure culture.

48 CONCLUSIONS: Reduction of TMAO by the gut microbiota (predominantly Enterobacteriaceae)

49 to TMA followed by host uptake of TMA into the bloodstream from the intestine and its conversion

50 back to TMAO by host hepatic enzymes is an example of metabolic retroconversion. TMAO

51 influences microbial metabolism depending on isolation source and taxon of gut bacterium.

52 Correlation of metabolomic and abundance data from mixed microbiota fermentation systems did not

53 give a true picture of which members of the gut microbiota were responsible for converting TMAO to

54 TMA; only by supplementing the study with pure culture work and additional metabolomics was it

55 possible to increase our understanding of TMAO bioconversions by the human gut microbiota.

56

2 bioRxiv preprint doi: https://doi.org/10.1101/225581; this version posted March 27, 2018. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.

57 BACKGROUND

58 Dietary methylamines such as choline, trimethylamine N-oxide (TMAO),

59 phosphatidylcholine (PC) and carnitine are present in a number of foodstuffs, including meat, fish,

60 nuts and eggs. It has long been known that gut bacteria are able to use choline in a fermentation-like

61 process, with trimethylamine (TMA), ethanol, acetate and ATP among the known main end-products

62 [1–3]. TMA can be used by members of the order Methanomassiliicoccales (Archaea) present in the

63 human gut to produce methane [4], or taken up by the host. Microbially produced TMA derived from

64 PC is absorbed in the small intestine [5]. TMA diffuses into the bloodstream from the intestine via the

65 hepatic vein to hepatocytes, where it is converted to trimethylamine N-oxide (TMAO) by hepatic

66 flavin-containing mono-oxygenases (FMOs) [6]. The bulk of TMAO, and lesser amounts of TMA,

67 derived from dietary methylamines can be detected in urine within 6 h of ingestion, and both

68 compounds are excreted in urine but not faeces [7,8]. TMAO can also be detected in human skeletal

69 muscle within 6 h of an oral dose of TMAO [8].

70 TMAO present in urine and plasma is considered a biomarker for non-alcoholic fatty liver

71 disease (NAFLD), insulin resistance and cardiovascular disease (CVD) [9–12]. Feeding TMAO to

72 high-fat-fed C57BL/6 mice exacerbates impaired glucose tolerance, though the effect on the gut

73 microbiota is unknown [13]. Low plasma PC and high urinary methylamines (including TMAO) were

74 observed in insulin-resistant mice on a high-fat diet, leading to the suggestion microbial methylamine

75 metabolism directly influences choline bioavailability; however, the microbiota of the mice was not

76 studied nor was labeled choline used to determine the metabolic fate of TMA/TMAO [9]. Choline

77 bioavailability is thought to contribute to hepatic steatosis in patients [11]. Spencer et al. [11]

78 manipulated dietary choline levels in 15 females for 2 weeks and monitored changes in the fecal

79 microbiota during the intervention. They found the level of choline supplementation affected

80 Gammaproteobacteria (inhibited by high levels of dietary choline, negatively correlated with changes

81 in liver fat/spleen fat ratios) and Erysipelotrichi (positively correlated with changes in liver fat/spleen

82 fat ratios), suggesting baseline levels of these taxa may predict the susceptibility of an individual to

83 liver disease. However, the study was under-powered and requires replication with a much larger

84 cohort to fully examine the link between choline bioavailability and liver disease. High levels of

3 bioRxiv preprint doi: https://doi.org/10.1101/225581; this version posted March 27, 2018. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.

85 circulating TMAO are associated with CVD [12], potentially via a platelet-mediated mechanism [14].

86 Circulating TMAO has been suggested to play a role in protection from hyperammonemia in mice,

87 acting as an osmoprotectant, and from glutamate neurotoxicity in mice and primary cultures of

88 neurons [15,16]. Recently, chronic exposure to TMAO was shown to attenuate diet-associated

89 impaired glucose tolerance in mice, and reduce endoplasmic reticulum stress and adipogenesis in

90 adipocytes [10]. The relevance of these findings to human health remains to be determined. There is a

91 basal level of TMA and TMAO detected in human urine even in the absence of dietary

92 supplementation [17], suggesting use of (microbial and/or host) cell-derived choline or PC in the

93 intestinal tract by the gut microbiota.

94 While it is well established that choline [2,11,18], PC [2,12] and carnitine [19–22] are used

95 by the human and rodent gut microbiotas to produce TMA, little is known about the reduction of

96 TMAO (predominantly from fish) to TMA (or other compounds) by members of these consortia.

97 TMAO is found at concentrations of 20–120 mg per 100 g fish fillet [23]. Ingestion of fish by humans

98 leads to increased urinary excretion of dimethylamine (DMA) and TMA, from 5.6 to 24.1 and from

99 0.2 to 1.6 µmol/24 h/kg of body weight, respectively [24]. Individuals with trimethylaminuria (fish

100 odour syndrome), in which TMA is not detoxified to TMAO by hepatic FMOs, have been shown to

101 reduce 40–60 % of an oral dose of TMAO to TMA [25]. It was suggested the gut microbiota was

102 responsible for reducing TMAO in these individuals, and in those individuals without

103 trimethylaminuria the TMA was re-oxidized in the liver before being excreted in the urine. The

104 process was termed ‘metabolic retroversion’ “(i.e., a cycle of reductive followed by oxidative

105 reactions to regenerate TMAO)” [25]. The reduction of TMAO to TMA is most commonly associated

106 with the gut microbiota of marine fish, with the TMA generated by these bacteria contributing to the

107 characteristic odour of rotting fish [26]. Members of the class Gammaproteobacteria, which includes

108 a number of food spoilage organisms, are known to reduce TMAO to TMA quantitatively and are

109 unable to reduce TMA further [26,27]. The conversion of TMAO to TMA by bacteria represents a

110 unique form of anaerobic respiration, in which TMAO reductase acts as a terminal electron acceptor

111 (from NADH or formate) for respiratory electron flow [27].

4 bioRxiv preprint doi: https://doi.org/10.1101/225581; this version posted March 27, 2018. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.

112 Mining of metagenomic data (torA) suggests Proteobacteria (particularly Escherichia and

113 Klebsiella spp.) are likely to contribute greatest to the production of TMA from TMAO in the human

114 gut via the TMAO reductase pathway, with (Eggerthellaceae) becoming more

115 important under stress [28]. Other microbial genes are associated with production of TMA from

116 choline (cutC), glycine-betaine (grdH), L-carnitine (cntA) and γ-butyrobetaine (cntA) [4]. A search of

117 the NCBI nucleotide database with the phrase ‘TMAO reductase’ also suggests many other bacteria of

118 human intestinal origin (namely, Salmonella, Helicobacter, Prevotella, Bacillus and Bacteroides spp.)

119 should be able to reduce TMAO to TMA. However, this trait has not been examined in vitro for

120 isolates of intestinal origin.

121 Consequently, we carried out an in vivo study in mice to confirm use of TMAO by the mouse

122 gut microbiota and to allow us to examine metabolic retroconversion of TMAO in a murine model.

123 We then used an in vitro fermentation system to highlight the effect of TMAO on the human faecal

124 microbiota. Finally, we screened a panel of human faecal and caecal/small intestinal isolates to

125 determine which members of the human gut microbiota were able to reduce TMAO to TMA, and

126 whether their metabolism was affected by being grown in the presence of TMAO.

127

128

129 METHODS

130 Animal work

131 Groups of six-week-old C57BL6/J mice (Janvier Labs, Courtaboeuf, France) were received

132 and acclimated in specific pathogen-free (SPF) maintenance conditions. They were fed with a

133 standard chow diet (R04-40, Safe, Augy, France) and were either given free access to tap water or tap

134 water containing the antibiotic cocktail (0.5 g/L vancomycin hydrochloride, 1 g/L neomycin trisulfate,

135 1 g/L metronidazole, 1 g/L ampicillin sodium; all antibiotics were purchased from Sigma-Aldrich) for

-4 136 14 days. Mice were given in the morning by gavage either a solution of d9-TMAO at 1×10 M

137 (Cambridge Isotope Laboratories Inc., DLM-4779-0, UK) or saline and euthanized 6 h later by

138 cervical dislocation. Blood samples were collected by tail tipping every 2 h in Microvette® CB 300

5 bioRxiv preprint doi: https://doi.org/10.1101/225581; this version posted March 27, 2018. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.

139 Lithium Heparin (Sarstedt, Marnay, France). Plasma was separated by centrifugation (10 min, 5000 g,

140 4 °C) and stored at -80 °C until analysed by ultra-performance liquid chromatography–tandem mass

141 spectrometry (UPLC–MS/MS).

142

143 UPLC–MS/MS determination of plasma TMA, d9-TMA, TMAO and d9-TMAO

144 UPLC–MS/MS was employed for the determination of TMA, d9-TMA, TMAO and d9-

13 15 145 TMAO. Samples (10 µL) were spiked with 10 µL Internal Standard solution (d9-choline and C3/ N-

146 TMA in water; 1 mg/L, Sigma-Aldrich). Ethyl 2-bromoacetate solution (45 µL) (15 g/L ethyl 2-

147 bromoacetate, 1 % NH4OH in acetonitrile; ChromaSolv grade, Sigma-Aldrich) was added and

13 15 148 derivatization of TMAs (TMA, d9-TMA and C3/ N-TMA) to their ethoxy-analogues was completed

149 after 30 min at room temperature. Protein/lipid precipitation solution (935 µL) (94 % acetonitrile/5

150 %water/1 % formic acid; ChromaSolv grade, Sigma-Aldrich) was added, samples were centrifuged

151 (20 min, 20,000 g, 4 °C) and were transferred to UPLC-autosampler vials. Sample injections (5 µL

152 loop) were performed with a Waters Acquity UPLC-Xevo TQ-S UPLC–MS/MS system equipped

153 with an Acquity BEH HILIC (2.1×100 mm, 1.7 µm) chromatographic column. An isocratic elution

154 was applied with 10 mM ammonium formate (Sigma-Aldrich) in 95:5 (v/v) acetronitrile:water for 7

155 min at 750 µL/min and 50 °C. Positive electrospray (ESI+) was used as ionization source and mass

156 spectrometer parameters were set as follows: capillary, cone and sources voltages at -700, -18 and

157 50 V, respectively, desolvation temperature at 600 °C, desolvation/cone/nebuliser gases were high

158 purity nitrogen (BOC) at 1000 L/h, 150 L/h and 7 bar, respectively. Collision gas was high-purity

159 argon (BOC). Mass spectrometer was operated in multiple reaction monitoring mode. The monitored

160 transitions were the following: for TMA, +146 → +118/59 m/z (23/27 V); for d9-TMA, +155 →

13 15 161 +127/68 m/z (21/23 V); for C3/ N-TMA, +150 → +63/122 m/z (27/22 V); for TMAO, +76 →

162 +59/58 m/z (12/13 V); for d9-TMAO, +85 → +68/66 m/z (18/20 V); and for d9-choline, +108 →

163 +60/45 m/z (20/22 V). The system was controlled by MassLynx software, also used for data

164 acquisition and analysis.

165

6 bioRxiv preprint doi: https://doi.org/10.1101/225581; this version posted March 27, 2018. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.

166 1H-NMR spectroscopy and data analysis

167 Medium samples were randomized and centrifuged (5 min, 16000 g). Aliquots (50 µL) of the

168 supernatants were diluted in 550 µL D2O containing 1 mM trimethylsilyl-(2,2,3,3-2H4)-1-propionate.

169 Samples were transferred to 5 mm NMR tubes and measured on a NMR spectrometer (Bruker)

170 operating at 600.22 MHz 1H frequency as described [29]. 1H-NMR spectra were pre-processed and

171 analysed as described [9] using the Statistical Recoupling of Variables-algorithm [30]. Structural

172 assignment was performed as described [31], using in-house and publicly available databases.

173

174 In vitro fermentation systems

175 Freshly voided faecal samples were obtained from three healthy human volunteers (one male

176 and two females; age range 29–31 years), none of whom had been prescribed antibiotics 6 months

177 prior to the study, eaten fish 3 days prior to sample collection or had any history of gastrointestinal

178 disease. The University of Reading’s Research Ethics Committee (UREC) does not require that

179 specific ethical review and approval be given by UREC for the collection of faecal samples from

180 healthy human volunteers to inoculate in vitro fermentation systems. Samples were processed

181 immediately by diluting them 1 in 10 (w/w) in pre-reduced minimal broth and homogenizing them in

182 a filter bag in a stomacher (Stomacher 400 Lab System; Seward) for 2 min at ‘high’ speed. For each

183 donor, a batch culture vessel containing 175 mL of pre-reduced minimal broth was inoculated with

184 20 mL of the faecal homogenate and 5 mL of sterile H2O containing 2 g TMAO dihydrate (Sigma-

185 Aldrich). Control vessels were inoculated with 20 mL of the faecal homogenate and 5 mL of sterile

186 H2O. The final working volume of each batch culture vessel was 200 mL. The pH of each vessel (pH

187 6.5) was controlled automatically by the addition of 2 M HCl or 2 M NaOH. pH controllers were

188 supplied by Electrolab. The contents of each vessel were stirred constantly. An anaerobic

189 environment was maintained by constantly sparging the vessels with O2-free N2. The temperature of

190 each vessel was maintained at 37 °C by use of a circulating waterbath connected to each fermentation

191 vessel. The experiment was run for 9 h, with samples taken at 0, 1, 2, 3, 4, 5, 6 and 9 h.

192

193 Fluorescence in situ hybridization (FISH) analysis

7 bioRxiv preprint doi: https://doi.org/10.1101/225581; this version posted March 27, 2018. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.

194 Aliquots (2× 375 µL) of sample were fixed in ice-cold 4 % paraformaldehyde for 4 h, washed

195 in sterile phosphate-buffered saline and stored for FISH analysis as described [32]. Supplementary

196 Table 1 gives details for probes used in this study. Probes were synthesized commercially (MWG-

197 Biotech) and labelled with the fluorescent dye cyanine 3 (Cy3; excitation λ, 514 nm; emission λ, 566

198 nm; fluorescence colour, orange–red). FISH was done as described [32].

199 Slides were viewed under a Nikon E400 Eclipse microscope. DAPI slides were visualized

200 with the aid of a DM400 filter; probe slides were visualized with the aid of a DM575 filter. Cells

201 (between 15 and 50 per field of view) were counted for 15 fields of view, and the numbers of bacteria

202 were determined by using the following equation:

203 DF × ACC × 6,732.42 × 50 × DFsample

204 Where the DF (dilution factor) was calculated by taking into account the concentration of the

205 original sample (375 µL to 300 µL = 0.8×). ACC (average cell count) was determined by counting 15

206 fields of view and assumes that a normal distribution was observed for the counts. The figure

207 6,732.42 refers to the area of the well divided by the area of the field of view. DFsample refers to the

208 dilution of sample used with a particular probe (e.g. 5× for t0 Lab158 counts). The detection limit of

209 this method was 89,766 bacteria/mL of sample (= log10 4.95).

210

211 Screening of bacteria for ability to reduce TMAO

212 Part of an in-house collection (University of Reading) of bacteria isolated from human

213 caecal/small intestinal and faecal samples (Table 1) was screened on minimal agar [g/L: glucose

214 (Fisher Scientific), 4; bacteriological peptone (Oxoid), 20; NaCl (Fisher Scientific), 5; neutral red

215 (Sigma-Aldrich), 0.03; agar technical no. 3 (Oxoid), 15; pH 7] with and without 1 % (w/v) TMAO

216 dihydrate (Sigma-Aldrich). The composition of the agar was based on [27], who used the medium

217 with crystal violet and bile salts to screen members of the Enterobacteriaceae for TMAO reductase

218 activity. Colonies of isolates able to ferment glucose were red, whereas those able to reduce TMAO

219 were white. Growth curves (OD600 measured hourly for 10–12 h, then at 24 h) were determined for

220 selected isolates grown anaerobically at 37 °C in anaerobic minimal broth [g/L: glucose, 4;

8 bioRxiv preprint doi: https://doi.org/10.1101/225581; this version posted March 27, 2018. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.

221 bacteriological peptone, 20; NaCl, 5; L-cysteine HCl (Sigma-Aldrich), 0.5; resazurin solution (0.25

222 mg/mL; Sigma-Aldrich), 4 mL; pH 7] with and without 1 % (w/v) TMAO dihydrate. Glucose was

223 substituted by raffinose (Sigma-Aldrich) [33] in the minimal broth when working with bifidobacteria

224 because of the poor growth of these bacteria on the glucose-based medium. pH and metabolite

225 profiles of culture medium were examined at the end of the growth experiment (i.e. at t24).

226

227 Statistical analyses

228 Differences between metabolites produced by faecal and caecal/small intestinal isolates of

229 Escherichia coli and lactic acid bacteria in the presence and absence of TMAO were analysed using

230 Student’s t test. FISH data from the in vitro fermentation systems were analysed using the

231 Kolmogorov–Smirnov test with statistical significance, after correction for multiple testing, taken at P

232 < 0.05. Because of the presence of ties in the metabolite data from in vitro fermentation systems, the

233 bootstrapped Kolmogorov–Smirnov test (10,000 replications) was used with these data. Data from in

234 vitro fermentation systems (FISH and NMR) were correlated using Spearman’s rank correlation

235 (corrected for ties) with results corrected for multiple testing using the method of Benjamini and

236 Hochberg [34].

237

238

239 RESULTS

240 In vivo confirmation of metabolic retroconversion

241 First, to confirm in vivo metabolic retroconversion (i.e. microbial conversion of TMAO to

242 TMA, followed by host conversion of TMA to TMAO), we administered isotopically-labelled d9-

243 TMAO or saline to mice that had or had not been treated with a broad-spectrum antibiotic cocktail for

244 14 days to suppress the gut microbiota [12]. Reduction of d9-TMAO to d9-TMA was quantified in

245 urine by UPLC–MS/MS up to 6 h after d9-TMAO gavage, together with unlabelled TMA and TMAO

246 potentially produced from dietary sources (Fig. 1).

247 In the absence of antibiotics, d9-TMAO was converted to d9-TMA within 2 h of gavage. This

248 conversion was dramatically reduced, leading to a significantly lower concentration of d9-TMA and a

9 bioRxiv preprint doi: https://doi.org/10.1101/225581; this version posted March 27, 2018. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.

249 higher concentration of d9-TMAO, when the gut microbiota was suppressed by antibiotics (Fig. 1A–

250 D). We noted that in the absence of antibiotics all animals excreted approximately three times higher

251 levels of urinary unlabelled TMA than TMAO (Fig. 1E–H), corresponding to constitutively low

252 FMO3 activity in mice [35]. Levels of unlabelled TMAO/TMA were not significantly different from

253 one another in the control and d9-TMAO-fed animals, while antibiotic treatment significantly reduced

254 the amount of both TMAO and TMA (Fig. 1E, F). Bioavailability of unlabelled TMAO/TMA, as

255 assessed by area under the curve (AUC), was significantly reduced by antibiotic treatment in both

256 saline and d9-TMAO-gavaged animals, with almost no TMA detected in either experimental group

257 (Fig. 1G, H).

258

259 Effect of TMAO on human gut bacteria within a mixed system

260 After in vivo validation of the role of the gut microbiota in metabolic retroconversion, we

261 analysed the effect of TMAO on the faecal microbiota in an anaerobic batch-culture fermentation

262 system. Fermenter vessels filled with the glucose-containing medium supplemented or not with 1 %

263 (w/v) TMAO were inoculated with faecal slurries from three healthy donors and monitored for 9 h.

264 With the exception of enhanced growth of the Enterobacteriaceae (probe Ent), the presence of

265 TMAO in the medium had no statistically significant effect on the growth of bacteria within the

266 fermentation systems at 9 h (Fig. 2A, Supplementary Figure 1).

267 Huge variability, as measured by 1H-NMR, was observed in the amount of TMA and DMA

268 produced by gut bacteria in the TMAO-containing fermentation systems, with the concentrations of

269 both metabolites increasing steadily from 0 to 9 h and differing significantly (P < 0.05) from the

270 control systems (Fig. 2B). The amount of TMAO in the systems was seen to decrease at 8 h.

271 Correlation of metabolite and FISH data demonstrated Clostridium clusters I and II

272 (Chis150), Enterobacteriaceae (Ent), bifidobacteria (Bif164) and coriobacteriia (Ato291) were

273 associated with TMA, acetate, ethanol and lactate (Fig. 2C). The Betaproteobacteria (Bet42a) were

274 anti-correlated with TMA, acetate, ethanol and lactate, which is unsurprising given this was the only

275 group of bacteria whose representation decreased in the fermentation systems over the course of the

276 experiment (Supplementary Figure 1, Fig. 2C). The Enterobacteriaceae and clostridia were

10 bioRxiv preprint doi: https://doi.org/10.1101/225581; this version posted March 27, 2018. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.

277 positively correlated with DMA production. The lactic acid bacteria (Lab158) were not significantly

278 correlated with any of the metabolites in the mixed culture.

279

280 Growth of pure cultures of gut bacteria in the presence of TMAO

281 Using an in-house collection of bacteria, we initially screened isolates on a modified version

282 of the agar of Takahi & Ishimoto [27] as a rapid means of screening bacteria for TMAO reductase

283 activity, and thereby their ability to reduce TMAO to TMA. While members of the

284 Enterobacteriaceae produced expected results (i.e. white colonies when grown in the presence of

285 TMAO and glucose, rather than red colonies on the glucose only control), we observed a number of

286 unexpected outcomes depending on the species under study. For example, colonies of lactic acid

287 bacteria were larger (almost twice their usual size) on TMAO-containing agar than on the glucose

288 control but remained red in colour, suggesting they had not reduced TMAO to TMA at detectable

289 levels but TMAO was influencing their growth. The clostridia examined produced mixed results on

290 the control and TMAO-containing media (i.e. white colonies on both plates or on the control plate

291 only, larger colonies on the TMAO-containing medium but without a colour change of the medium).

292 To determine whether these isolates were converting TMAO to TMA but a low level, we

293 examined the growth of all isolates in liquid culture and metabolites in the spent medium using NMR.

294 The growth of the Enterobacteriaceae was most greatly affected by the presence of TMAO in the

295 medium, with a faster, longer-lasting exponential phase than for the same isolates grown in the

296 control medium (Fig. 3A). pH of the spent medium (after 24 h) when Enterobacteriaceae were grown

297 in the presence of TMAO increased from a mean of 4.7 ± 0.3 (for the control) to 7.6 ± 0.3 (n = 20)

298 (the change in pH is what causes the colonies to appear white on TMAO-containing agar). The

299 growth of lactic acid bacteria, including Enterococcus and Streptococcus (Fig. 3A) spp., was

300 enhanced in the presence of TMAO, but not to the same extent as seen for the Enterobacteriaceae.

301 There was no significant difference (P = 0.27, t test) in the pH of the spent medium for these bacteria

302 after 24 h (mean 4.67 ± 0.9 compared with 4.33 ± 0.27 for the control). The growth of members of

303 Clostridium cluster I (e.g. Clostridium perfringens, Fig. 3A) was not enhanced in the presence of

304 TMAO, though some of these bacteria changed the colour of both the control and TMAO-containing

11 bioRxiv preprint doi: https://doi.org/10.1101/225581; this version posted March 27, 2018. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.

305 plates yellow during their growth. The pH of the spent liquid medium confirmed this observation to

306 be due to the alkalinity of the media in the control and TMAO-containing media after 24 h incubation:

307 e.g. Clostridium sporogenes D1(9) (pH 6.25 compared with 6.76 in the control medium), Clostridium

308 paraputrificum L16-FAA6 (pH 6.45 vs 5.38) and Clostridium perfringens L20-BSM1 (pH 5.56 vs

309 4.64).

310 1H-NMR analysis of spent medium from the TMAO-containing and control samples

311 demonstrated that, as expected, the Enterobacteriaceae produced the greatest amount of TMA from

312 TMAO (mean 38.79 ± 11.08 mM compared with 0.03 ± 0.01 mM, n = 20) (Fig. 3B; Supplementary

313 Table 2). Members of the families Peptostreptococcaceae (3.72 mM, n = 1), Clostridiaceae (Cluster

314 I) (2.62 ± 1.83 mM, n = 3), Porphyromonadaceae (1.42 mM, n = 1), Bacteroidaceae (1.40 ± 0.31

315 mM, n = 3), Enterococcaceae (1.19 ± 0.05 mM, n = 5), Erysipelotrichaceae (0.94 mM, n = 1;

316 [Clostridium] ramosum), Staphylococcaceae (0.34 mM, n = 1), Streptococcaceae (0.30 ± 0.16 mM, n

317 = 5), Lactobacillaceae (0.17 ± 0.07 mM, n = 2), Pseudomonadaceae (0.12 mM, n = 1) and

318 (0.13 ± 0.1 mM, n = 17) produced low levels of TMA from TMAO (Fig. 3B;

319 Supplementary Table 2). There was great variability in the ability of the bifidobacteria to produce

320 TMA from TMAO, with several isolates and [Clostridium] innocuum, Actinomyces odontolyticus,

321 Fusobacterium ulcerans and Actinomyces viscosus not producing TMA from TMAO (Fig. 3B;

322 Supplementary Table 2).

323

324 Differences in the metabolic capabilities of faecal and caecal Escherichia coli isolates

325 Comparison of the amounts of TMA, and co-metabolites, produced by the faecal (n = 7) and

326 caecal (n = 9) isolates of Escherichia coli demonstrated significantly higher amounts of TMA were

327 produced by the caecal isolates compared with the faecal isolates in the TMAO-containing medium

328 (Fig. 3C). Escherichia coli of caecal origin produced more TMA than faecal isolates of the same

329 bacterium or other enterobacteria (Hafnia, Citrobacter and Klebsiella spp.) (Supplementary Table

330 2). The faecal isolates produced more acetate and lactate than the caecal isolates when grown in the

331 control medium. Taken together, these results demonstrate the different metabolic capabilities of

332 isolates of Escherichia coli recovered from different regions of the human gut.

12 bioRxiv preprint doi: https://doi.org/10.1101/225581; this version posted March 27, 2018. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.

333

334 Lactic acid bacteria produce more lactate in the presence of TMAO

335 Differences were also seen in the amount of lactic acid produced by lactic acid bacteria in the

336 presence and absence of TMAO (Fig. 3B, D). In raffinose-containing medium, bifidobacteria

337 produced increased of amounts of lactate when grown in the presence of TMAO (the bifidobacteria

338 grew poorly, if at all, in glucose-containing media). Unlike the bifidobacteria, the Streptococcaceae

339 and Enterococcaceae grew well in the glucose-containing medium and produced over 25 mM lactic

340 acid in the TMAO-containing samples compared with <5 mM in the control samples (Fig. 3D). To the

341 best of our knowledge, this is the first time TMAO has been shown to influence the metabolism of gut

342 bacteria – specifically lactic acid bacteria – without producing appreciable amounts of TMA.

343

344

345 DISCUSSION

346 TMAO is a circulating metabolite produced as a direct result of microbial degradation of

347 dietary methylamines in the intestinal tract, and can be readily detected along with its precursor TMA

348 in human urine, blood and skeletal muscle [8,12,18,22]. It worsens atherosclerosis in some mouse

349 models of CVD, and is positively correlated with CVD severity in humans. Beneficial effects

350 associated with TMAO include potential protection from hyperammonia and glutamate neurotoxicity,

351 alleviation of endoplasmic reticulum stress and improved glucose homeostasis by stimulating insulin

352 secretion by pancreatic cells [10,15,16].

353 Over many decades it has been established that choline, PC and carnitine are dietary

354 methylamines that contribute directly to microbiome-associated circulating levels of TMAO found in

355 humans and other animals – e.g. [2,12,18,19,21,22]. However, TMAO itself is a water-soluble

356 osmolyte found in high abundance in fish. Based on their observations that individuals with

357 trimethylaminuria could reduce an oral dose of TMAO to TMA could not detoxify it to TMAO, al-

358 Waiz et al. [25] suggested the gut microbiota could use TMAO as a substrate, and in those individuals

359 without trimethylaminuria the TMA was re-oxidized in the liver before being excreted in the urine.

360 This led these authors to propose the process of ‘metabolic retroversion’. Reference to the literature

13 bioRxiv preprint doi: https://doi.org/10.1101/225581; this version posted March 27, 2018. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.

361 associated with fish spoilage and recent mining of metagenomic data have predicted members of the

362 Enterobacteriaceae (particularly Escherichia coli and Klebsiella pneumoniae) have the potential to

363 convert TMAO to TMA in the intestinal tract, but this has not been tested in vitro or in vivo to date

364 [4,8,26–28]. Consequently, we instigated this study to demonstrate metabolic retroconversion of

365 TMAO, and to determine the effect of TMAO on the growth and metabolism of human-derived

366 intestinal bacteria in pure and mixed cultures.

367 Through in vivo administration of deuterated TMAO to mice via oral gavage in mice, we

368 unambiguously demonstrated that TMAO is converted to TMA, with this TMA detectable in plasma

369 within 2 h of administration. This conversion was highly dependent on the gut microbiota, as

370 conversion of TMAO to TMA was dramatically reduced when the microbiota was suppressed by

371 treatment of animals with broad-spectrum antibiotics. Even in the presence of antibiotics, there was

372 low-level conversion of TMAO to TMA, suggesting a subpopulation of the microbiota was resistant

373 to the antibiotics used in our experiment. However, administration of a broad-spectrum antibiotic

374 cocktail for 14 days has been demonstrated to be an effective means of suppressing the gut microbiota

375 in studies associated with gut microbial use of dietary methylamines [12].

376 Our in vivo experiment clearly demonstrates the gut microbiota converts TMAO to TMA, and

377 that this TMA is re-oxidized to TMAO, in line with the process of ‘metabolic retroversion’ [25]. Gut-

378 associated microbial conversion of the majority of TMAO to TMA is at odds with recent findings [8],

379 in which it was suggested TMAO is taken up intact and not metabolised by the gut microbiota of

380 humans. The ratio of TMAO-to-TMA is around 10:1 in humans and 1:10 in mice, meaning TMAO

381 was N-oxidized before Taesuwan et al. [8] were able to observe it in circulating blood of their human

382 subjects. In a human system with high FMO3 activity, 17O-labelled TMAO would need to be used in

383 any study evaluating O turnover to allow calculation of the true rate of retroconversion.

384 Based on work done in mice [10,36], metabolic retroconversion of TMAO may be protective,

385 and may even go some way to explaining why TMA and TMAO are detected at low levels in urine in

386 the absence of dietary methylamines [17]. Chronic exposure of high-fat-fed mice to TMAO reduced

387 diet-associated endoplasmic stress and adipogenesis, and improved glucose tolerance [10]. Exposure

388 of high-fat-fed mice to TMA reduced low-grade inflammation and insulin resistance via inhibition of

14 bioRxiv preprint doi: https://doi.org/10.1101/225581; this version posted March 27, 2018. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.

389 interleukin-1 receptor-associated kinase 4 (IRAK-4) [36]. Concentrations of the microbial signalling

390 metabolite TMA used by Chilloux et al. [36] were comparable to those found in normal human

391 plasma. Continual exposure of the human system to TMA/TMAO may bring similar benefits to host

392 metabolic and inflammatory responses; these remain to be studied.

393 Having conducted our in vivo experiment in mice, we examined the ability of a range of

394 human-derived gut bacteria to convert TMAO to TMA. In the mixed microbiota system and in pure

395 cultures, the growth of the Enterobacteriaceae – the main TMA-producers – was quickly affected by

396 the presence of TMAO. This is likely to happen in the human gut also. Consequently, we believe

397 dietary TMAO undergoes metabolic retroconversion in , with the TMA produced as a result

398 of bacterial activity in the gut available to the host for conversion back to TMAO by FMOs in

399 hepatocytes. It should be noted that [8] did not suppress or monitor the intestinal/faecal microbiota

400 when they administered isotopically labeled TMAO to humans, nor did they measure d9-TMA and d9-

401 TMAO in the portal vein, bypassing subsequent hepatic N-oxidation of d9-TMA, so it is not possible

402 to interpret their results in the context of presence/absence of microbial activity.

403 Of note is the finding that caecal isolates of enterobacteria produce more TMA from TMAO

404 than faecal isolates of the same bacterium. The speed with which TMAO was reduced to TMA by the

405 Enterobacteriaceae in the present study suggests bacterial conversion of TMAO to TMA takes place

406 in the small intestine/proximal colon of humans and small intestine/caecum of mice. It is, therefore,

407 unsurprising that caecal bacteria – representing the microbiota present at the intersection of the small

408 and – are metabolically more active than their faecal counterparts with respect to

409 TMAO metabolism. This finding is relevant to functional studies of the gut microbiota where gene

410 annotations are based largely on faecal isolates, whose functionalities may be greatly different from

411 those of bacteria in other regions of the human intestinal tract [37,38]. It has already been

412 demonstrated that the microbiota of the small intestine is enriched for functions associated with rapid

413 uptake and fermentation of simple carbohydrates compared with the faecal microbiota, and that

414 streptococci isolated from this niche are functionally very different from the same bacteria isolated

415 from different habitats [37,38]. It is, therefore, important we characterize the functions and genomes

416 of bacteria isolated from all regions of the intestinal tract, not just those of faecal bacteria, to gain a

15 bioRxiv preprint doi: https://doi.org/10.1101/225581; this version posted March 27, 2018. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.

417 true picture of how microbial activity influences host health.

418 Enterobacteriaceae made the greatest contribution to the conversion of TMAO to TMA, both

419 in pure culture and in a mixed microbiota. These Gram-negative bacteria are a source of the virulence

420 factor lipopolysaccharide (LPS), which is associated with low-grade inflammation in high-fat-fed

421 mice and elevated plasma levels that define metabolic endotoxaemia [39,40]. High-fat feeding has

422 been shown to increase the representation of Enterobacteriaceae in the caecal microbiota of obesity-

423 prone Sprague–Dawley rats [41], though this mode of feeding is known to modulate the microbiota of

424 mice independent of obesity [42]. Non-LPS-associated virulence of Enterobacteriaceae, Vibrio

425 cholerae and Helicobacter pylori is increased when these bacteria are grown anaerobically or

426 microaerophilically in the presence of as little as 5–10 mM TMAO [43–46], and may be an additional

427 means by which the gut microbiota contributes to CVD and other diseases in which increased

428 representation of Gammaproteobacteria is observed. This warrants attention in future animal studies.

429 Lactic acid bacteria clearly grow better in the presence of TMAO. The relatively high

430 concentration (1 %) of TMAO used in this study may have contributed to this improved growth, as

431 TMAO is an osmolyte that stabilizes proteins. Future work will involve growing lactic acid bacteria in

432 a range of TMAO concentrations to determine how this compound affects their growth and gene

433 expression, and comparing faecal and caecal isolates. Similar to the Enterobacteriaceae, a large

434 number of these bacteria are facultative anaerobes able to grow over a range of conditions, and whose

435 representation is increased in obese, and cirrhotic patients [47,48]. Streptococcus and Enterococcus

436 spp. are commensal lactic acid bacteria of the gut microbiota known to modulate immune function;

437 but little is known about their metabolic activities in mixed microbial populations [49]. Understanding

438 how commensal lactic acid bacteria influence the host in dysbiosis in mixed microbial communities

439 may allow the development of approaches to modulate their activity and influence host health.

440 With respect to the lactic acid bacteria, it is important to note that our mixed culture work did

441 not highlight these as being relevant to TMAO metabolism. This is unsurprising given these bacteria

442 do not produce large quantities of TMA from TMAO. However, we have shown their metabolism is

443 affected by presence of TMAO in growth medium, and the increased lactate they produce in its

444 presence may contribute to cross-feeding associated with short-chain fatty acid production [50]. It is

16 bioRxiv preprint doi: https://doi.org/10.1101/225581; this version posted March 27, 2018. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.

445 difficult to determine relevance of correlations from mixed microbial ecosystems in the absence of

446 isotope labelling or pure culture work: i.e. correlation does not equate with causation. As an example,

447 three (Clostridium clusters I and II, bifidobacteria, coriobacteriia) of the four groups of bacteria

448 correlated with TMA production in our fermentation study did not produce notable quantities of TMA

449 from TMAO based on our pure culture work. Therefore, correlating microbiota and metabolite data

450 derived from complex systems will not give a true picture of which members of microbiota contribute

451 to specific metabolic processes, and work with pure cultures is required to supplement functional

452 studies to increase our understanding of poorly understood microbially driven metabolic processes

453 within the human gut.

454

455

456 CONCLUSIONS

457 We have demonstrated metabolic retroconversion – another example of host–microbial co-

458 metabolism – occurs in the mammalian system with respect to TMAO, whereby TMAO is reduced by

459 the gut microbiota to TMA and regenerated by host hepatic enzymes. We have also demonstrated that

460 growth and metabolism of members of the gut microbiota are affected by TMAO in a source- and

461 taxon-dependent manner, with the family Enterobacteriaceae making the greatest contribution to

462 production of TMA in the gut.

463

464

465 ABBREVIATIONS

466 AUC, area under the curve; DMA, dimethylamine; FISH, fluorescence in situ hybridization; FMO,

467 flavin mono-oxygenase; MMA, monomethylamine; PC, phosphatidylcholine; TMA, trimethylamine;

468 TMAO, trimethylamine N-oxide; UPLC–MS/MS, ultra-performance liquid chromatography–tandem

469 mass spectrometry.

470

471

472 DECLARATIONS

17 bioRxiv preprint doi: https://doi.org/10.1101/225581; this version posted March 27, 2018. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.

473 Ethics approval: All animal procedures were authorized following review by the institutional ethics

474 committee (Sorbonne Universities) and carried out under national license conditions. Ethical approval

475 to collect caecal effluent from patients was obtained from St Thomas’ Hospital Research Ethics

476 Committee (06/Q0702/74) covering Guy’s and St Thomas’ Hospitals, and transferred by agreement to

477 London Bridge Hospital. Patients provided written consent to provide samples.

478 Consent for publication: Not applicable.

479 Availability of data and material: The datasets used and/or analysed during the current study are

480 available from the corresponding authors upon reasonable request.

481 Competing interests: The authors have no competing interests.

482 Funding: MLJP was sponsored by Fundación Alfonso Martin Escudero (Spain). Work was supported

483 by a grant from the EU-FP7 (METACARDIS, HEALTH-F4-2012-305312) to DG and MED. LH is in

484 receipt of an MRC Intermediate Research Fellowship in Data Science (grant number MR/L01632X/1,

485 UK Med-Bio).

486 Authors’ contributions: MLJP, LH and ALM performed the microbiological work, JC was responsible

487 for NMR analysis and interpretation of the NMR data, TA and FB did the animal work, MED

488 supervised the NMR work, ALM supervised the microbiological work, CM provided invaluable

489 advice for the labelling study and DG supervised the animal work. All authors contributed to the

490 writing of the manuscript.

491 Acknowledgements: Not applicable.

492

493 494 REFERENCES 495 [1] Wünsche R. Uber die wirkung von darmbakterien auf trimethylammoniumbasen. Ber Ges 496 Physiol Exp Pharm. 1940;118:303. 497 [2] Zeisel SH, Wishnok JS, Blusztajn JK. Formation of methylamines from ingested choline and 498 lecithin. J Pharmacol Exp Ther. 1983;225:320-24. 499 [3] Bradbeer C. The clostridial fermentations of choline and ethanolamine. 1. Preparation and 500 properties of cell-free extracts. J Biol Chem. 1965;240:4669-74.

18 bioRxiv preprint doi: https://doi.org/10.1101/225581; this version posted March 27, 2018. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.

501 [4] Borrel G, McCann A, Deane J, Neto MC, Lynch DB, Brugère JF, et al. Genomics and 502 metagenomics of trimethylamine-utilizing Archaea in the human gut microbiome. ISME J. 503 2017;11:2059-74. 504 [5] Stremmel W, Schmidt KV, Schuhmann V, Kratzer F, Garbade SF, Langhans CD, et al. Blood 505 trimethylamine-N-oxide originates from microbiota mediated breakdown of phosphatidylcholine 506 and absorption from small intestine. PLoS One. 2017;12:e0170742. 507 [6] Bennett BJ, de Aguiar Vallim TQ, Wang Z, Shih DM, Meng Y, Gregory J, et al. 508 Trimethylamine-N-oxide, a metabolite associated with atherosclerosis, exhibits complex genetic 509 and dietary regulation. Cell Metab. 2013;17:49-60. 510 [7] de la Huerga J, Popper H. Urinary excretion of choline metabolites following choline 511 administration in normals and patients with hepatobiliary diseases. J Clin Invest. 1951;30:463- 512 70. 513 [8] Taesuwan S, Cho CE, Malysheva OV, Bender E, King JH, Yan J, et al. The metabolic fate of 514 isotopically labeled trimethylamine-N-oxide (TMAO) in humans. J Nutr Biochem. 2017;45:77- 515 82. 516 [9] Dumas ME, Barton RH, Toye A, Cloarec O, Blancher C, Rothwell A, et al. Metabolic profiling 517 reveals a contribution of gut microbiota to fatty liver phenotype in insulin-resistant mice. Proc 518 Natl Acad Sci U S A. 2006;103:12511-6. 519 [10] Dumas ME, Rothwell AR, Hoyles L, Aranias T, Chilloux J, Calderari S, et al. Microbial-host 520 co-metabolites are prodromal markers predicting phenotypic heterogeneity in behavior, obesity, 521 and impaired glucose tolerance. Cell Rep. 2017;20:136-48. 522 [11] Spencer MD, Hamp TJ, Reid RW, Fischer LM, Zeisel SH, Fodor AA. Association between 523 composition of the human gastrointestinal microbiome and development of fatty liver with 524 choline deficiency. Gastroenterology. 2011;140:976-86. 525 [12] Tang WH, Wang Z, Levison BS, Koeth RA, Britt EB, Fu X, et al. Intestinal microbial 526 metabolism of phosphatidylcholine and cardiovascular risk. N Engl J Med. 2013;368:1575-84. 527 [13] Gao X, Liu X, Xu J, Xue C, Xue Y, Wang Y. Dietary trimethylamine N-oxide exacerbates 528 impaired glucose tolerance in mice fed a high fat diet. J Biosci Bioeng. 2014;118:476-81. 529 [14] Zhu W, Wang Z, Tang WHW, Hazen SL. Gut microbe-generated trimethylamine N-oxide from 530 dietary choline is prothrombotic in subjects. Circulation. 2017;135:1671-3. 531 [15] Kloiber O, Banjac B, Drewes LR. Protection against acute hyperammonemia: the role of 532 quaternary amines. Toxicology. 1988;49:83-90. 533 [16] Miñana MD, Hermenegildo C, Llsansola M, Montoliu C, Grisolía S, Felipo V. Carnitine and 534 choline derivatives containing a trimethylamine group prevent ammonia toxicity in mice and 535 glutamate toxicity in primary cultures of neurons. J Pharmacol Exp Ther. 1996;279:194-9. 536 [17] de la Huerga J, Gyorgy P, Waldstein S, Katz R, Popper H. The effects of antimicrobial agents 537 upon choline degradation in the intestinal tract. J Clin Invest. 1953;32:1117-20.

19 bioRxiv preprint doi: https://doi.org/10.1101/225581; this version posted March 27, 2018. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.

538 [18] Wang Z, Klipfell E, Bennett BJ, Koeth R, Levison BS, Dugar B, et al. Gut flora metabolism of 539 phosphatidylcholine promotes cardiovascular disease. Nature. 2011;472:57-63. 540 [19] Rebouche CJ, Mack DL, Edmonson PF. L-Carnitine dissimilation in the gastrointestinal tract of 541 the rat. Biochemistry. 1984;23:6422-6. 542 [20] Rebouche CJ, Chenard CA. Metabolic fate of dietary carnitine in human adults: identification 543 and quantification of urinary and fecal metabolites. J Nutr. 1991;121:539-46. 544 [21] Seim H, Schulze J, Strack E. Catabolic pathways for high-dosed L(-)- or D(+)-carnitine in germ- 545 free rats? Biol Chem Hoppe Seyler. 1985;366:1017-21. 546 [22] Koeth RA, Wang Z, Levison BS, Buffa JA, Org E, Sheehy BT, et al. Intestinal microbiota 547 metabolism of L-carnitine, a nutrient in red meat, promotes atherosclerosis. Nat Med. 2013;19, 548 576-85. 549 [23] Jebsen JW, Riaz M. Breakdown products of trimethylamine oxide in airdried stockfish. Means 550 of enhancing the formation of formaldehyde and dimethylamine. Fish Dir Skr Ser Ernoering. 551 1977;1:145-53. 552 [24] Zeisel SH, Da Costa KA. Increase in human exposure to methylamine precursors of N- 553 nitrosamine after eating fish. Cancer Res. 1986;46:6136-8. 554 [25] al-Waiz M, Ayesh R, Mitchell SC, Idle JR, Smith RL. Disclosure of the metabolic retroversion 555 of trimethylamine N-oxide in humans: a pharmacogenetic approach. Clin Pharmacol Ther. 556 1987;42:608-12. 557 [26] Barrett EL, Kwan HS. Bacterial reduction of trimethylamine oxide. Ann Rev Microbiol. 558 1985;39:131-49. 559 [27] Takagi M, Ishimoto M. Escherichia coli mutants defective in trimethylamine N-oxide reductase. 560 FEMS Microbiol Lett. 1983;17:247-50. 561 [28] Jameson E, Doxey AC, Airs R, Purdy KJ, Murrell JC, Chen Y. Metagenomic data-mining 562 reveals contrasting microbial populations responsible for trimethylamine formation in human 563 gut and marine ecosystems. Microb Genom. 2016;2:e000080. 564 [29] Dona AC, Jiménez B, Schäfer H, Humpfer E, Spraul M, Lewis MR., et al. Precision high- 565 throughput proton NMR spectroscopy of human urine, serum, and plasma for large-scale 566 metabolic phenotyping. Anal Chem. 2014;86:9887-94. 567 [30] Blaise BJ, Shintu L, Elena B, Emsley L, Dumas ME, Toulhoat P. Statistical recoupling prior to 568 significance testing in nuclear magnetic resonance based metabonomics. Anal Chem. 569 2009;81:6242-51. 570 [31] Dona AC, Kyriakides M, Scott F, Shephard EA, Varshavi D, Veselkov K, et al. A guide to the 571 identification of metabolites in NMR-based metabonomics/metabolomics experiments. Comput 572 Struct Biotechnol J. 2016;14:135-53.

20 bioRxiv preprint doi: https://doi.org/10.1101/225581; this version posted March 27, 2018. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.

573 [32] Martín-Peláez S, Gibson GR, Martín-Orúe SM, Klinder A, Rastall RA, La Ragione RM, et al. In 574 vitro fermentation of carbohydrates by porcine faecal inocula and their influence on Salmonella 575 Typhimurium growth in batch culture systems. FEMS Microbiol Ecol. 2008;66:608-19. 576 [33] Trojanová I, Vlková E, Rada V, Marounek M. Different utilization of glucose and raffinose in 577 breve and Bifidobacterium animalis. Folia Microbiol (Praha). 2006;51:320-4. 578 [34] Benjamini Y, Hochberg Y. Controlling the false discovery rate: a practical and powerful 579 approach to multiple testing. J R Stat Soc Ser B. 1995;57:289-300. 580 [35] Zhang J, Cerny MA, Lawson M, Mosadeghi R, Cashman JR. Functional activity of the mouse 581 flavin-containing monooxygenase forms 1, 3, and 5. J Biochem Mol Toxicol. 2007;21:206-15. 582 [36] Chilloux J, Everard A, Zhang L, Blancher C, Fuchs J, Hoyles L, et al. Gut microbial 583 trimethylamine inhibits IRAK-4 and chronic inflammation in metabolic syndrome. bioRxiv. 584 2018;doi:10.1101/277434. 585 [37] Zoetendal EG, Raes J, van den Bogert B, Arumugam M, Booijink CC, Troost FJ, et al. The 586 human small intestinal microbiota is driven by rapid uptake and conversion of simple 587 carbohydrates. ISME J. 2012;6:1415-26. 588 [38] van den Bogert B, Boekhorst J, Herrmann R, Smid EJ, Zoetendal EG, Kleerebezem M. 589 Comparative genomics analysis of Streptococcus isolates from the human small intestine reveals 590 their adaptation to a highly dynamic ecosystem. PLoS One. 2013;8:e83418. 591 [39] Cani PD, Amar J, Iglesias MA, Poggi M, Knauf C, Bastelica D, et al. Metabolic endotoxemia 592 initiates obesity and insulin resistance. Diabetes. 2007;56:1761-72. 593 [40] Cani PD, Bibiloni R, Knauf C, Waget A, Neyrinck AM, Delzenne NM, et al. Changes in gut 594 microbiota control metabolic endotoxemia-induced inflammation in high-fat diet-induced 595 obesity and diabetes in mice. Diabetes. 2008;57:1470-81. 596 [41] de La Serre CB, Ellis CL, Lee J, Hartman AL, Rutledge JC, Raybould HE. Propensity to high- 597 fat diet-induced obesity in rats is associated with changes in the gut microbiota and gut 598 inflammation. Am J Physiol Gastrointest Liver Physiol. 2010;299:G440-8. 599 [42] Hildebrandt MA, Hoffmann C, Sherrill-Mix SA, Keilbaugh SA, Hamady M, Chen YY, et al. 600 High-fat diet determines the composition of the murine gut microbiome independently of 601 obesity. Gastroenterology. 2009;137:1716-24. 602 [43] Ando H, Abe H, Sugimoto N, Tobe T. Maturation of functional type III secretion machinery by 603 activation of anaerobic respiration in enterohaemorrhagic Escherichia coli. Microbiology. 604 2007;153:464-73. 605 [44] Lee KM, Park Y, Bari W, Yoon MY, Go J, Kim SC, Lee HI, et al. Activation of cholera toxin 606 production by anaerobic respiration of trimethylamine N-oxide in Vibrio cholerae. J Biol Chem. 607 2012;287:39742-52.

21 bioRxiv preprint doi: https://doi.org/10.1101/225581; this version posted March 27, 2018. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.

608 [45] Paiva JB, Penha Filho RA, Pereira EA, Lemos, MV, Barrow PA, Lovell MA, et al. The 609 contribution of genes required for anaerobic respiration to the virulence of Salmonella enterica 610 serovar Gallinarum for chickens. Braz J Microbiol. 2009;40:994–1001. 611 [46] Wu D, Cao M, Peng J, Li N, Yi S, Song L, et al. The effect of trimethylamine N-oxide on 612 Helicobacter pylori-induced changes of immunoinflammatory genes expression in gastric 613 epithelial cells. Int Immunopharmacol. 2017;43:172-8. 614 [47] Qin N, Yang F, Li A, Prifti E, Chen Y, Shao L, et al. Alterations of the human gut microbiome 615 in liver cirrhosis. Nature. 2014;513:59-64. 616 [48] Le Chatelier E, Nielsen T, Qin J, Prifti E, Hildebrand F, Falony G, et al. Richness of human gut 617 microbiome correlates with metabolic markers. Nature. 2013;500:541-6. 618 [49] van den Bogert B, Meijerink M, Zoetendal EG, Wells JM, Kleerebezem M. Immunomodulatory 619 properties of Streptococcus and Veillonella isolates from the human small intestine microbiota. 620 PLoS One. 2014;9:e114277. 621 [50] Hoyles L, Wallace RJ. Gastrointestinal tract: intestinal fatty acid metabolism and implications 622 for health. In Handbook of Hydrocarbon and Lipid Microbiology, vol. 4, part 7, Human-Animal- 623 Plant Health and Physiology Consequences of Microbial Interactions with Hydrocarbons and 624 Lipids. 2010, 3120-32. Edited by T. McGenity, K. N. Timmis, J. R. van der Meer & V. de 625 Lorenzo. New York: Springer. 626 [51] Roger LC, Costabile A, Holland DT, Hoyles L, McCartney AL. Examination of faecal 627 Bifidobacterium populations in breast- and formula-fed infants during the first 18 months of life. 628 Microbiology. 2010;156:3329-41. 629 [52] Hoyles L, Murphy J, Neve H, Heller KJ, Turton JF, Mahony J, et al. Klebsiella pneumoniae 630 subsp. pneumoniae–bacteriophage combination from the caecal effluent of a healthy woman. 631 PeerJ. 2015;doi:10.7717/peerj.1061. 632

22 bioRxiv preprint doi: https://doi.org/10.1101/225581; this version posted March 27, 2018. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.

633 FIGURE LEGENDS

634 Fig. 1. In vivo confirmation of metabolic retroconversion of TMAO. Reduction of d9-TMAO to d9-

635 TMA was quantified by UPLC–MS/MS up to 6 h after d9-TMAO gavage and antibiotic treatment,

636 together with unlabelled TMA and TMAO levels. Plasma quantification of post-gavage (A) d9-TMA

637 and (B) d9-TMAO. *, Significantly (P < 0.05; t test and corrected for multiple comparison using the

638 Holm–Sidak method) different from the respective groups not treated with antibiotics. (C) d9-TMA

639 bioavailability (AUC). (D) d9-TMAO bioavailability (AUC). Plasma quantification of post-gavage

640 unlabelled/endogenous (E) TMA and (F) TMAO. *, Significant between d9 and d9 antibiotic

641 treatment; $, significant between TMAO and TMAO antibiotic treatment. (G) Unlabelled/endogenous

642 TMA bioavailability (AUC). (H) Unlabelled/endogenous TMAO bioavailability (AUC). Data (n = 6

643 per group) are shown as mean ± SEM. (A, B, E and F). Differences between the bioavailabilities (C,

644 D, G and H) were assessed using one-way analysis of variance (ANOVA), followed by Holm–Sidak

645 post hoc tests. Data with different superscript letters are significantly different (P < 0.05).

646

647 Fig. 2. Effect of TMAO on mixed faecal microbial population in vitro. (A) Enumeration of selected

648 bacteria in fermentation vessels by FISH analysis. Red lines, TMAO-containing systems; blue lines,

649 negative controls. Data are shown as mean + SD (n = 3). Eub338, total bacteria; Ent,

650 Enterobacteriaceae; Bif164, Bifidobacterium spp.; Lab158, lactic acid bacteria. *, Statistically

651 significantly different (adjusted P < 0.05) from the control at the same time point. Full data are shown

1 652 in Supplementary Figure 1. (B) H-NMR data for batch culture samples. Data are shown mean ± SD

653 (n = 3). Red lines, TMAO-containing systems; blue lines, negative controls. *, Statistically

654 significantly different (P < 0.05) from the negative control at the same time point. (C) Bidirectional

655 clustering of correlation matrix of FISH data and data for the six metabolites found in highest

656 amounts in the NMR spectra from the batch-culture samples. +, Adjusted P value (Benjamini–

657 Hochberg) statistically significant (P < 0.05). FISH and metabolite data and a table of correlations and

658 adjusted P values (Benjamini–Hochberg) for the batch-culture samples are available in

659 Supplementary Tables 3–5.

660

23 bioRxiv preprint doi: https://doi.org/10.1101/225581; this version posted March 27, 2018. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.

661 Fig. 3. Influence of TMAO on growth and metabolism of pure cultures of gut bacteria. (A)

662 Representative growth curves for isolates grown in the presence and absence of TMAO. Red lines,

663 TMAO-supplemented cultures; blue lines, negative controls. Data are shown as mean ± SD (n = 3).

664 (B) Biplot showing production of various metabolites when isolates were grown in the presence of

665 TMAO. Summary of data from Supplementary Table 2. The larger a circle, the more of the

666 metabolite produced by an isolate. (C) Differences in metabolites produced when caecal and faecal

667 isolates of Escherichia coli were grown in the presence (+) and absence (–) of 1 % TMAO. Adjusted

668 (Benjamini–Hochberg) P values indicate the caecal isolates were significantly different from the

669 faecal isolates for a particular metabolite. (D) Lactate production by lactic acid bacteria was increased

670 in the presence of TMAO. Enterobacteriaceae, n = 20; Bifidobacteriaceae, n = 17; Streptococcaceae,

671 n = 7; Enterococcaceae, n = 5. Members of the Enterococcaceae and Streptococcaceae are

672 homofermenters (produce only lactic acid from glucose fermentation), whereas the

673 Bifidobacteriaceae are heterofermenters (produce ethanol, CO2 and lactic acid from glucose

674 fermentation), though it should be noted the bifidobacteria included in this study were grown on

675 raffinose-containing media. Red, TMAO-containing medium; blue, negative control. *, Statistically

676 significantly different from its negative control (adjusted P value < 0.05).

24 bioRxiv preprint doi: https://doi.org/10.1101/225581; this version posted March 27, 2018. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.

677 Fig. 1

678

25 bioRxiv preprint doi: https://doi.org/10.1101/225581; this version posted March 27, 2018. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.

679 Fig. 2

680

26 bioRxiv preprint doi: https://doi.org/10.1101/225581; this version posted March 27, 2018. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.

681 Fig. 3

682

27 bioRxiv preprint doi: https://doi.org/10.1101/225581; this version posted March 27, 2018. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.

683 Table 1. Details for human-derived gut bacteria screened for their ability to reduce or utilize TMAO

Isolate* Identified as Source Reference Facultative anaerobe? D2(9) [Clostridium] innocuum Human faeces M.L. Jiménez-Pranteda, unpublished No L16-FAA1 [Clostridium] ramosum Human caecum L. Hoyles, unpublished No L12-BSM1 Actinomyces odontolyticus Human caecum L. Hoyles, unpublished No L6-BSM10 Actinomyces viscosus Human caecum L. Hoyles, unpublished No L6-FAA7 Bacteroides fragilis Human caecum L. Hoyles, unpublished No D1(4) Bacteroides vulgatus Human faeces M.L. Jiménez-Pranteda, unpublished No L9-FAA7 Bacteroides vulgatus Human caecum L. Hoyles, unpublished No DSM 20083T Bifidobacterium adolescentis Human intestine DSMZ† No LCR26 Bifidobacterium animalis subsp. lactis Infant faeces [51] No LCR11 Bifidobacterium bidifum Infant faeces [51] No L25-MRS2 Bifidobacterium bifidum Human caecum L. Hoyles, unpublished No LCR5 Bifidobacterium breve Infant faeces [51] No LCR8 Bifidobacterium breve Infant faeces [51] No LCR1 Bifidobacterium dentium Infant faeces [51] No DSM 20093T Bifidobacterium gallicum Human intestine DSMZ No LCR6 Bifidobacterium longum Infant faeces [51] No DSM 20088T Bifidobacterium longum subsp. infantis Infant intestine DSMZ No LCR2 Bifidobacterium longum subsp. infantis Infant faeces [51] No DSM 20219T Bifidobacterium longum subsp. longum Human intestine DSMZ No LCR3 Bifidobacterium pseudocatenulatum Infant faeces [51] No L19-MRS1 Bifidobacteroum longum subsp. longum Human caecum L. Hoyles, unpublished No L25-MRS8 Bifidobacteroum longum subsp. longum Human caecum L. Hoyles, unpublished No L26-MRS4 Bifidobactrium animalis subsp. lactis Human caecum L. Hoyles, unpublished No L26-FAA1 Citrobacter gillenii Human caecum L. Hoyles, unpublished Yes L8-FAA3 Citrobacter koseri Human caecum L. Hoyles, unpublished Yes L16-FAA6 Clostridium paraputrificum Human caecum L. Hoyles, unpublished No L20-BSM1 Clostridium perfringens Human caecum L. Hoyles, unpublished No D1(9) Clostridium sporogenes Human faeces M.L. Jiménez-Pranteda, unpublished No D2(14) Enterococcus faecalis Human faeces M.L. Jiménez-Pranteda, unpublished Yes D3(1) Enterococcus faecalis Human faeces M.L. Jiménez-Pranteda, unpublished Yes D5(2) Enterococcus faecalis Human faeces M.L. Jiménez-Pranteda, unpublished Yes D6(1) Enterococcus faecium Human faeces M.L. Jiménez-Pranteda, unpublished Yes D6(5) Enterococcus gallinarum Human faeces M.L. Jiménez-Pranteda, unpublished Yes D1(2) Escherichia coli Human faeces M.L. Jiménez-Pranteda, unpublished Yes D2(1) Escherichia coli Human faeces M.L. Jiménez-Pranteda, unpublished Yes D2(2) Escherichia coli Human faeces M.L. Jiménez-Pranteda, unpublished Yes D2(8) Escherichia coli Human faeces M.L. Jiménez-Pranteda, unpublished Yes D3(8) Escherichia coli Human faeces M.L. Jiménez-Pranteda, unpublished Yes D4(15) Escherichia coli Human faeces M.L. Jiménez-Pranteda, unpublished Yes D5(1) Escherichia coli Human faeces M.L. Jiménez-Pranteda, unpublished Yes L1-FAA5 Escherichia coli Human caecum L. Hoyles, unpublished Yes L13-FAA2 Escherichia coli Human caecum L. Hoyles, unpublished Yes L16-FAA5 Escherichia coli Human caecum L. Hoyles, unpublished Yes L19-FAA2 Escherichia coli Human caecum L. Hoyles, unpublished Yes L20-FAA3 Escherichia coli Human caecum L. Hoyles, unpublished Yes L24-FAA5 Escherichia coli Human caecum L. Hoyles, unpublished Yes L5-FAA2 Escherichia coli Human caecum L. Hoyles, unpublished Yes L6-FAA1 Escherichia coli Human caecum L. Hoyles, unpublished Yes L9-MRS1 Escherichia coli Human caecum L. Hoyles, unpublished Yes L9-FAA5 Fusobacterium ulcerans Human caecum L. Hoyles, unpublished Yes L15-FAA9 Hafnia paralvei Human caecum L. Hoyles, unpublished No L4-FAA5 Klebsiella pneumoniae subsp. pneumoniae Human caecum [52] Yes L26-MRS5 fermentum Human caecum L. Hoyles, unpublished No

28 bioRxiv preprint doi: https://doi.org/10.1101/225581; this version posted March 27, 2018. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.

L26-FAA6 Lactobacillus rhamnosus Human caecum L. Hoyles, unpublished No L13-FAA10 Parabacteroides johnsonii Human caecum L. Hoyles, unpublished No D4(1) Paraclostridium bifermentans Human faeces M.L. Jiménez-Pranteda, unpublished No L1-FAA6 Pseudomonas aeruginosa Human caecum L. Hoyles, unpublished Yes D2(4) Staphyloccoccus hominis Human faeces M.L. Jiménez-Pranteda, unpublished Yes D1(5) Streptococcus anginosus Human faeces M.L. Jiménez-Pranteda, unpublished Yes D5(12) Streptococcus anginosus Human faeces M.L. Jiménez-Pranteda, unpublished Yes L25-MRS1 Streptococcus gallolyticus Human caecum L. Hoyles, unpublished Yes L4-MRS5 Streptococcus oralis Human caecum L. Hoyles, unpublished Yes L4-MRS1 Streptococcus sanguinis Human caecum L. Hoyles, unpublished Yes D4(3) Streptococcus sp. Human faeces M.L. Jiménez-Pranteda, unpublished Yes L26-MRS7 Streptococcus vestibularis Human caecum L. Hoyles, unpublished Yes LCR4 Unknown Bifidobacterium Infant faeces [51] No 684

685 *Ln- prefix, different numbers indicate isolates recovered from different individuals.

686 †DSMZ, Leibniz Institute DSMZ – German Collection of Microorganisms and Cell Cultures.

29